Our objectives are (1) to account for aphasic deficits in terms of psycholinguistic theory and (2) to test models of language processing by extending them to the phenomena of language breakdown. Section (1) concerns studies of lexical-phonological retrieval: to amass evidence on individual differences in error patterns in naming and repetition, including changes linked to recovery; to strengthen evidence that nonword errors generated in naming are influenced by lexical factors; to analyze how structural and sequential factors influence nonword errors in naming; and to determine whether movement errors in connected speech occur at above chance rates in individual aphasics. (2) concerns context-sensitive naming disorders. The aims are to develop reliable diagnostic procedures for these disorders; and to test and extend the theoretical accounts proposed by investigators who reported them. (3 ) presents studies of sentence processing that focus on aphasic deficits and grammatical processing. Aims include: to evaluate "mapping" and "working memory" accounts of agrammatic comprehension; to determine whether the poor performance of Wernicke's aphasics on grammaticality judgment tests reflects deficits at the single word level or grammatical structures; to replicate with new methods the finding that complexity of meaning facilitates verb access in some patients; and to provide data to test a computational account of this phenomenon that relates it to agrammatism. (4) links computational modeling to the empirical work described above: to compare two models of lexical retrieval for their ability to capture naming, repetition, and recovery data; to extend these models to account for neighborhood density effects, context-sensitive naming disorders and other challenging data; to develop a new connectionist learning model that produces multisyllabic words as true sequences and test this model against nonword phonological errors; and to develop a model of grammatical encoding that can be used to understand impaired access to verbs and multiple word sequences.